The CUbRIK Fashion Trend Analysis Application: presented to CUbRIK Summer School - July 2014 - by Claudio Massari (Innovation Engineering)

1. CUbRIK SummerSchool2014
CUbRIK Summer School 0
CUbRIK Fashion Trends
Bringing innovation to SMEs in the Fashion market
Claudio Massari –Innovation Engineering srl

2. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 1
Agenda
Innovation Engineering
The Fashion domain
The rationale behind the application
The CUbRIK approach
Main challenges
Conclusion

3. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 2
Innovation Engineering
Private Italian SME
Specialisationin the design & development of advanced IT solutionssupporting knowledge extractionand management within organizations
Main Clients: Large Corporation, SMEs, Public bodies;
Involved in several R&D European and National projects

4. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 3
The Fashion domain
The Fashion domain
Fashion is a multi-billion Euro industry
representing about 3% of the EU GDP
90% of the organisationsare SMEs
The fashion domain is characterized by:
Strong investments in new processes, new materials, new products
High seasonability: quick adaptation to the market requirements

5. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 4
The rationale behind CUbRIK Fashion Trends
SMEs needs
SMEs are constantly seeking for innovative
products and new trends
Data for Trends Analysis are collected manually:
surveys, interviews
Expensive Analysisperformed by specialized agencies
and consultants
Gathering of the data is expensive
AnalisingTrends is expensive

6. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 5
SMEs’ needs
Why doesn’t your company carry out trend analysis when planning new products

7. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 6
The rationale
Aim of the CUbRIKFashion Trend
Supporting European SMEs in the fashion sector to:
Exploit the potential of the new technology
Spot trends as soon as they evolve
To identifypotential customer preferences
Get feedback from and learn about the needs of fashion consumers.
To trigger new trends through open innovation with SME’s own target groups

8. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 7
CUbRIK Fashion Trend main features
Trend analysis
What may SMEs obtain?
Trend analysis based on:
Garment category
A sample image
Time and spatial constraints
Results obtained:
The most popular colors, textures
An insight on the most popular images
The best combination of colors
Need addressed:
Affordable Trend analysis
Spot of the most popular colors
for a garment

9. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 8
CUbRIK Fashion Trend main features
Create an outfit and ask feedback
What may SMEs obtain?
Create and test an outfit based on:
A sample image
A set of similar images
Time and spatial constraints
Results obtained:
Creating an outfit for further analysis
Feedback about the outfit from their
network
Need addressed:
Affordable perception of the customers feeling
Support for realising collections (producing, buying, distributing)

10. CUbRIK SummerSchool2014
Ask for feedback
SMEs obtain to:
Test a garment based on:
A sample image
Results obtained:
Receiving feedback about the
garment from their network
Need addressed:
Testing new product
Affordable perception of the customers feeling
Support for realising collections
(producing buying, distributing)
2-4/07/2014 CUbRIK Summer School 9
CUbRIK Fashion Trend main features

11. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 10
The CUbRIK pillars
Human in the loop
Gatheringusers’ generated content
Adopting a mixed solution: automatic analysis and Human in the loop approach
User preferences identification of the from the Social Network
Results:
Detecting trends on user preferences basis
Multimedia search similarity supporting SMEs in feeling the market

12. CUbRIK SummerSchool2014
Usage of mixed crowdsourcing resources:
Crowdworkers
Players of Games With A Purpose –GWAP
Social Network users (preferences) and SMEs’ network (feedback)
Rationale
The “human touch” enhances the results and carries out key tasks
The “human feeling” iscrucial in the fashion sector:
SMEs require analysis based on human feeling
2-4/07/2014 CUbRIK Summer School 11
The CUbRIK Human in the loop approach

13. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 12
Gathering images
Usage of the user’s content:
Social networks are a tangible resource: users generate large amount of data
High usage of the Social Network and the games: in 2011, they were the 1stand 2ndonline activities, surpassing email [Nielsen]
200k images are crawled each week in Twitter
Main issue
Mixed quality of the images
Affordable huge quantity of data for SMEs

14. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 13
The CUbRIK mixed approach
Software analysis + Human in the loop provide SMEs with:
Better Image identification
Is this Fashion? Automatic crawling is not capable to answer
Enhancement of results
More precise feature analysis results
available for the Trend detection

15. CUbRIK Summer School 2014
2-4/07/2014 CUbRIK Summer School 14
The CUbRIK mixed approach
Just humans can correctly understand similarity:
 Are these images similar?
 The human perception of the similarity «dimensions» provides
SMEs users with similarity point of view
 Skirt length
 Pattern
 Cut
 Trendiness
 Casualness
 Suitability for a type of weather
SMEs are provided with results enhanced by
humans
Vergleichen von Kleidungsstücken
Ähneln sich die Röcke (Farbe,
Schnitt, Länge, …)?

16. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 15
Gathering preferences
Users preferences on Twitter (based on users activities)
Users preferences from videos
(which are the preferred
keyframes?)
Gathering feedback from yourcustomers’ network
Do you like this image or this outfit?
SMEs can rely on users/customerspreferences

17. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 16
The main challenge
The main challenges in the Fashion trends detecting:
Generation of the dataset (identification of fashion images)
Identification of the garments
Analysismultimedia features (color, texture)
Popularitydetection
Similarity identification
The CUbRIKApproach:
Crawlingand human in the loop for filtering fashion images
Automatic and manual (GWAP) segmentations of the garments
Automatic processing of images (features extraction)
Popularity identified through users activities in Social networks
Human in the loop for similarity perception identification

18. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 17
Benefits
Innovation potentials
Innovative tool providing SMEs with a affordable market trend analysis
User “fresh” based generated contents. Affordable
Easy detection of popular behaviours/trendsand colour/texture by location/season
Identification of similarities and recommendations provision
SMEs (small actors in the fashion sector, B2C):
Get over budget constraints for wide market analysis
Switchfrom an analysisbased on inner personnel experience to an affordable analysis based on their potential customers and trends
Large organizations (buyers, distributors, producers addressing general public)
Tackle the need of detect and pilot trends obtaining affordable results in order to manage production, buying and distribution
Obtain an additional/alternative tool to retrieve trend analysis;

19. CUbRIK SummerSchool2014
2-4/07/2014 CUbRIK Summer School 18
THANKS!